Beruflich Dokumente
Kultur Dokumente
COMPUTING
LECTURE 1
Computation may someday be organized as a public utility.
- John McCarthy, 1961
2
CLOUD COMPUTING
3
INTRODUCTION
CLOUD DEFINITION
4
INTRODUCTION
CLOUD DEFINITION
The U.S. National Institute of Standards and Technology (NIST) denes cloud
computing as:
Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access
to a shared pool of congurable computing resources (e.g., networks, servers, storage,
applications, and services) that can be rapidly provisioned and released with minimal
management effort or service provider interaction.
5
INTRODUCTION
CLOUD ARCHITECTURES
Scalable resource allocation
Tailored services
Software as a Service (SaaS)
Platform as a Service (PaaS)
Infrastructure as a Service (IaaS)
6
INTRODUCTION
CLOUD ARCHITECTURES
7
ACHITECTURE: LAYERED CLOUD MODEL
INTRODUCTION
CLOUD COMPUTING LAYERS
Application Service (SaaS)
MS Live/Exchange, Google Docs, Salesforce.com, Quicken Online, Jupyter
Application Platform (PaaS)
Google App Engine, Heroku, AWS
Server Platform (IaaS)
Google Compute Engine, Amazon EC2, OpenStack, Eucalpytus
9
INTRODUCTION
CLOUD COMPUTING LAYERS
Services Description
Services Complete business services such as PayPal,
Services OpenID, OAuth, Google Maps, Alexa
Capacity
Resources
Resources
Demand Capacity
Demand
Time Time
Unused resources
14
CHARACTERISTICS OF CLOUD COMPUTING
On-demand self service:
Cloud computing resources can be provisioned on-demand by the users, without
requiring interactions with the cloud service provider. The process of
provisioning resources is automated.
smartphones.
CHARACTERISTICS OF CLOUD COMPUTING
Resource pooling:
The computing and storage resources provided by cloud service providers are
pooled to serve multiple users using multi-tenancy. Multi-tenant aspects of the
cloud allow multiple users to be served by the same physical hardware.
Rapid elasticity:
Cloud computing resources can be provisioned rapidly and elastically. Cloud
resources can be rapidly scaled up or down based on demand.
16
CHARACTERISTICS OF CLOUD COMPUTING
Reliability:
Applications deployed in cloud computing environments generally have a higher reliability
since the underlying IT infrastructure is professionally managed by the cloud service.
Multi-tenancy:
The multi-tenanted approach of the cloud allows multiple users to make use of the same
shared resources.
In virtual multi-tenancy, computing and storage resources are shared among multiple users.
In organic multi-tenancy every component in the system architecture is shared among multiple
tenants
17
CLIENT/SERVER VS. CLOUD ARCHITECTURE
Cloud Cloud
Interface Admin
Storage Server
Switch/
Router
Network Compute
Network
Node
Compute
Node
Storage
Client Client Client Client Client Client Node
19
TYPES OF CLOUD
Public clouds
by independent service providers
Users have concerns on data security and privacy
Private clouds
Not much different from traditional internal computing clusters
Typically used by big companies
Hybrid clouds
One part is private; the other is public
Address the concerns on data security and privacy
Virtual Private Clouds
Provided by public cloud providers
Using VPN to isolate from the public cloud
INTRODUCTION
SUPER CLOUDS
21
INTRODUCTION
22
WHAT MOTIVATED CLOUD COMPUTING
Initial motivation:
Web-scale problems
Solutions:
Large data centers
How to access:
Highly-interactive Web applications (thin client)
23
INITIAL MOTIVATION: WEB-SCALE PROBLEMS
Characteristics:
Definitely data-intensive
May also be processing intensive
Examples:
Crawling, indexing, searching, mining the Web
Post-genomics life sciences research
Other scientific data (physics, astronomers, etc.)
Sensor networks
Web 2.0 applications
SmartThings/home integration 24
HOW MUCH DATA?
Google processes over 24 PB a day (24k terabytes)
CERNs LHC generates 25 PB a year
all words ever spoken by human beings ~ 5 EB (5m terabytes)
Amount of data that exists in the digital universe 3+ ZB (3b terabytes)
Brain Research through Advancing Innovative Neurotechnologies (BRAIN) project est:
multiple yottabyes (trillions of terabytes)
LARGE data is the next frontier
How do we store this amount of data?
HDD density
SDD density 25
26
CLOUDS
Allow access to applications other than on local computer or internet connected
device
But
Only as long as have internet connection
Sensitive information
How much do you trust the public cloud vendor?
Application integration (exchange info when local and on cloud)
28
CLOUD COMPONENTS
3 components
Clients
Datacenter
Distributed servers
29
CLOUD COMPONENTS
Clients
Mobile
SmartPhones, Tablets, Service Hubs
Thin
no internal hard drives, lets servers do all work, displays info
Thick
Laptops, desktop computers
Which is the best?
Thin - lower costs, security, power consumption, easy to replace, less noise
30
DATA CENTER
Data Center facility used to house computer systems and
associated components
31
DISTRIBUTED SERVERS
Servers host the resources needed by cloud users
Compute nodes
Provides CPU, Memory, Scratch Storage, and Networking resources through virtualized interfaces.
Hosts guest operating systems (Virtual Machines) using one or more VM hypervisors
Resource interface depends on the type of cloud (horizontal/vertical cloud)
Storage nodes
Compute nodes only provide temporary storage space for users/applications
Storage nodes provide long term data storage solutions
Can be mapped to specific processes running on compute nodes, users, interface applications, etc.
Administrative nodes
Provides hidden back-end services such as resource load balancing, administrative/resource
32
33
IMPROVEMENTS SINCE 80S
Disk capacity
From 10s MB to several TB orders of magnitude
IBM built 120PB storage array
Bandwidth
1-10Mbps to 100s of Gbps
34
IMPROVEMENTS SINCE 80S
CPU Improvements:
Transistor shrink, increased clock rate, advanced instruction pipeline, cache
memory, multicores, faster bus interconnections, lower energy consumption
https://www.top500.org/lists/ - lists the top 500 highest (known)
performance super clusters
Current leader: Sunway TaihuLight (China)
10.6 million CPU cores
1.3PB RAM
Power consumed while fully operational:
15.3mW
Linpack: 93PFlop/s (.093EFlop/s) or
35
93,000,000,000,000,000 floating point operations per second.
By comparison, your laptop: ~125 Gflop/s, Smartphone: ~400 Mflop/s
SOLUTION FOR COMPUTATIONAL GROWTH:
LARGE DATA CENTERS
Web-scale problems? Throw more machines at it!
Decades ago computing power in mainframes in computer rooms
Personal computers changed that
Now, network data centers with centralized computing are back in vogue
In the future businesses will not need to invest in a data center
How can we easily access datacenter resources to fit our needs?
36